OFT Basic tuning procedure

[HSayaovong]

Originally Posted by phrosty

Excellent. I've made an app to automate it for me -- even better than the linked howto because it properly interpolates between bins. If anyone is interested I can make it a little more user-friendly and throw it up on a webpage.

Please do. Would be useful

[phrosty]

Originally Posted by shif7i7down

Sounds like something that can be created in excel, and all you had to do was paste in your logg files.

Which would be awesome if someone with the excel technical know how could make one. And share it on here

Sadly though I am a programmer, I possess only minimal Excel-fu.

[Kodename47]

Originally Posted by shif7i7down

Sounds like something that can be created in excel, and all you had to do was paste in your logg files.

Which would be awesome if someone with the excel technical know how could make one. And share it on here

You might want to look through this:
http://www.romraider.com/forum/viewforum.php?f=32

[Td-d]

Vgi has actually already put together a programme to handle both open and closed loop based on the numerous spreadsheets out - I've test driven it, works perfectly and lightening fast. Excel is a dog for doing these kind of number crunching exercises - especially if you're generating logs with the OP2 which will be large.

See here: http://www.romraider.com/forum/viewt...p?f=32&t=10481

[Kodename47]

I'm testing it out too, needs some wrinkles sorting but I like it. Nice to ser tools being developed. I just wish I could programme to help out.

[Lytheum]

Originally Posted by Td-d

Vgi has actually already put together a programme to handle both open and closed loop based on the numerous spreadsheets out - I've test driven it, works perfectly and lightening fast. Excel is a dog for doing these kind of number crunching exercises - especially if you're generating logs with the OP2 which will be large.

See here: http://www.romraider.com/forum/viewt...p?f=32&t=10481

got it to work by installing 64 bit java, but now when i open rom raider it says requires 32 bit java and some features may be unavailable. doh!

[vgi]

Originally Posted by Lytheum

got it to work by installing 64 bit java, but now when i open rom raider it says requires 32 bit java and some features may be unavailable. doh!

Let me see what I built the jar file with. But if you know how to rebuild it with your java version (as long as it's v1.6+) - the source code is there provided for public.

And yes, thanks to Kodename47's suggestions and spreadsheets I will be making some changes to make it more user friendly (especially for ecutek folks) and adding new views.

PS you don't need 64bit java, you just need v1.6+ java

[solidONE]

Originally Posted by vgi

Let me see what I built the jar file with. But if you know how to rebuild it with your java version (as long as it's v1.6+) - the source code is there provided for public.

And yes, thanks to Kodename47's suggestions and spreadsheets I will be making some changes to make it more user friendly (especially for ecutek folks) and adding new views.

PS you don't need 64bit java, you just need v1.6+ java

I used your app using the CL portion of the app and over did the smoothing. Threw my MAF scaling way off. I tried a second time without smoothing and LTFT's hover around -3% to +2%. What would you recommend to getting the LTFT's even tighter?


BTW how do you get the AFR% error from the logs? Specifically from an OFT log.

[Kodename47]

Originally Posted by solidONE

I used your app using the CL portion of the app and over did the smoothing. Threw my MAF scaling way off. I tried a second time without smoothing and LTFT's hover around -3% to +2%. What would you recommend to getting the LTFT's even tighter?


BTW how do you get the AFR% error from the logs? Specifically from an OFT log.

How much data are you collecting? CL requires a fair bit of data to work well. I'm pushing @vgi for some graphical views to see what's happening. The spreadsheets that this is based on only deal with 1 form of injection. There could be more at play than just MAF scaling. Also be aware that where the CL portion of the MAF scaling ends that the data collection will be minimal. It's easy to skew the scale if the process is 100% automated.

[solidONE]

Originally Posted by Kodename47

How much data are you collecting? CL requires a fair bit of data to work well. I'm pushing @vgi for some graphical views to see what's happening. The spreadsheets that this is based on only deal with 1 form of injection. There could be more at play than just MAF scaling. Also be aware that where the CL portion of the MAF scaling ends that the data collection will be minimal. It's easy to skew the scale if the process is 100% automated.

I inputted about at least an half hour's worth of mostly deliberate smooth throttle input for the CL. For the OL I did 7 3rd gear pulls and applied the LTFT to the AFR % error. Could not copy and paste the maf scale into the romraider table, so I had to do it one by one. Just loaded the new modified rom with the OL changes. Will see how it goes tomorrow.

I also converted the 7 pulls into dynographs It seems I lost the few ponies I gained with the messed up maf scale, from about 196HPS back down to about 189HPS... Not sure what to make of it. I'm still hitting all the additional KC ignition advance I added from before.

Log of the pulls: http://www.datazap.me/u/solidone/e85...g=0&data=1-5-6

[Kodename47]

Do you have excel? An easy way to get the RR copy to work is to copy it into excel, move the MAFv row to the right one cell and then input [Table2D] into the 1st cell. Then you can copy that into RR. As ECUtek uses a vertical table, that makes it slightly more awkward again and I've asked this to be looked at. Maybe @vgi can add a RomRaider copy button to do that automatically?

I wouldn't add the LTFT to the % error, you want the output AFR to be corrected 1st. Although the end difference would be minimal, it's just not quite as precise. Again, this has been raised

[Lytheum]

Originally Posted by vgi

Let me see what I built the jar file with. But if you know how to rebuild it with your java version (as long as it's v1.6+) - the source code is there provided for public.

And yes, thanks to Kodename47's suggestions and spreadsheets I will be making some changes to make it more user friendly (especially for ecutek folks) and adding new views.

PS you don't need 64bit java, you just need v1.6+ java

Weird....says can't find javaw when I try to run the batch file.

[vgi]

Originally Posted by solidONE

I used your app using the CL portion of the app and over did the smoothing. Threw my MAF scaling way off. I tried a second time without smoothing and LTFT's hover around -3% to +2%. What would you recommend to getting the LTFT's even tighter?


BTW how do you get the AFR% error from the logs? Specifically from an OFT log.

You should take a look at the Usage tab. The scaling is meant to be run few times (with new maf scaling each time and new logs) until you don't see anymore improvements (afr %error, there is scattered plot checkbox). (@Kodename47 I think this is why calcing mean in CL should be just fine as well as after few runs the results should be very similar).

Also, in Usage tab I believe I describe how the 'Smoothing' view works. The curve you see in the 'Smoothing' view is NOT a MAF curve, it's the smoothing slope curve. To see the smoothed curve (after you have applied let's say default 3 degree) you need to tick off 'smoothing' checkbox which will bring you back to results view and tick on 'corrected' and 'smoothed' checkboxes - then you will see raw corrected maf curve vs smoothed maf curve.

[vgi]

Originally Posted by Lytheum

Weird....says can't find javaw when I try to run the batch file.

I'm not sure what you have installed but you could change 'javaw' to just 'java', you will only see the command line window in addition to the app window.